CMPE 344 Computer Networks

Spring 2024

Internetworking
Part 2
Reading: Peterson and Davie, §3.3, 3.5, 4.1, 4.2

Sources of slides:
Peterson and Davie, Computer Networks: A Systems Approach, 6th ed., Morgan Kaufmann, 2021.
Tanenbaum, Feamster, and Wetherall, Computer Networks, 6th ed., Pearson, 2021.
Kurose and Ross, Computer Networking, 8th ed., Pearson, 2021.

23/02/2024 1
 Aim and Problems
• Aim: Build networks connecting millions of users around
the globe spanning networks based on any technology
• Problems: heterogeneity and scalability
– heterogeneity: need to support different LANs, point-to-
point technologies, switched networks, different
addressing formats
– scalability: addressing (management and
configuration) and routing must be able to handle
millions of hosts
• We will examine the (original) IP protocol (IPv4), IP
addressing, packet forwarding, subnetting, and IPv6.

2
 Outline
• Internet architecture
• IP service model
• IP forwarding
• Address translation (ARP)
• Automatic host configuration (DHCP) and
error reporting (ICMP)
• Virtual Private Networks (VPNs)
• Subnetting
• Supernetting: Classless routing (CIDR)
• IPv6

3
 The Internet and IP
Terminology

– network = network based on one technology

– internet = “network of networks”
–The Internet = internet using IP
– routers = nodes connecting networks
– IP = Internet Protocol, current version IPv4 (IP
Version 4)

4
 Internetworking

• Internetwork: An arbitrary collection of networks

interconnected to provide some sort of host-host to
packet delivery service
5
 IP principles
• Each host (each network interface)
– has a globally unique IP address
– must be reachable by everyone and from
everywhere
• Simple packet forwarding
– Network nodes simply forward packets
– Keeping the routers as simple as possible
was one of the original design goals of IP

6
 Internet architecture
• Recall Internet architecture from Chapter 1
– Application layer: FTP, HTTP, …(Last Chapter)
– Transport layer: TCP (reliable byte transfer) and UDP (unreliable
datagram delivery) provide logical channels to applications (Next
Chapter)
– Network or IP layer: IP protocol interconnects multiple networks
into a single logical network (This Chapter)
– “Link” layer: wide variety of LAN and point-to-point protocols

FTP HTTP NV TFTP

TCP UDP

IP

■■■
NET1 NET2 NETn 7
Protocol stack

8
 IP service model

• Main idea in the Internet service model:

– Make it undemanding enough that IP can be
run over anything
• This model is the major reason for the success
of IP technology
• Service model consists of 2 parts:
– Data delivery model
– Global addressing scheme

9
 Data delivery model

• Data delivery in the Internet

– IP network connectionless (datagram-based)
– IP network offers best-effort delivery (unreliable
service)
• packets may be lost
• packets may be delivered out of order
• duplicate copies of a packet may be delivered
• packets can be delayed for a long time
• “intelligence” implemented at the end hosts
– datagram format (next slide)

10
 IP datagram format
Format aligned at 32 bit words
– simplifies packet processing in software

Header fields
– Version: version 4 (IPv4) or version 6 (IPv6)
– HLen: header length, in 32-bit words (min 5)
– TOS: type of service, used to give
priorities to packets (QoS issue)
– Length: data+header length, in bytes
– 2nd word for fragmentation/reassembly (next
slide)
– TTL: time to live, no of times packet allowed
to be forwarded (no of hops), default 64,
detects packets caught in routing loop
– Protocol: identifies upper layer protocols,
TCP (6), UDP (17)
– Checksum: erroneous packets discarded
– Addresses: global Internet addresses
– Options: rarely used
11
 Checksum vs. CRC
• Checksum is calculated by considering the
entire IP header as a sequence of 16-bit words,
adding them up using ones’ complement
arithmetic, and taking the ones’ complement of
the result
• If any bit in the header is corrupted in transit, the
checksum will not contain the correct value upon
receipt of the packet
• Checksuming does not have the same strong
error detection properties as a CRC, but it is
much easier to calculate in software
12
 Fragmentation/reassembly
• Each network has an MTU (Maximum Transmission Unit)
– e.g., Ethernet 1500 bytes, PPP 532 bytes
• Strategy
– Fragment when necessary (MTU < datagram length)
– In general, try to avoid fragmentation: Hosts are encouraged to
perform “path MTU discovery”
• Fragments are self-contained datagrams
– each fragment contains a common identifier in Ident field
– Flags (M-bit) and Offset used to guide fragmentation process
– Offset measured in 8 byte units
– Fragmented packet can be again re-fragmented
• Reassembly is performed only at destination host
• Reassembly does not try to recover lost fragments

13
 Fragmentation/reassembly example
• Assume H5 wants to send a 1400 byte datagram to H8
(see slide 8): 1400 bytes + 20 bytes (IP header)
• MTU: 802.11 and Ethernet 1500 bytes, PPP 532 bytes

376 376

14
 Example (continued)

(a) Unfragmented packet: H5R1R2

(b) Fragmented packets:

R2R3H8

Can you think of a strategy

that will abuse fragmentation
to carry out a denial-of-service
attack?
How can one prevent such
an attack?
15
 IP global addressing
• Properties
– globally unique, 32 bits
– hierarchical:
network part + host part
– address identifies interface
• end host has one interface
• router has many interfaces
– IP address  domain name
• Original classful addressing
– class A, B and C networks
– defines different sized networks
– idea: small no of WANs, modest no of A : 1.0.0.0 to 127.255.255.255
campus networks, large no of LANs B : 128.0.0.0 to 191.255.255.255
• Dotted Decimal Notation C : 192.0.0.0 to 223.255.255.255
– 32 bit addresses represented as
groups of 8 bit integers
– e.g., 12.34.158.5

16
 Forwarding vs. routing

• Forwarding:
– Process of taking a packet from input interface,
looking at its destination address, consulting a
forwarding table, and sending the packet over an
output interface determined by that table

• Routing:
– Process by which forwarding tables are constructed
– Routing must create optimal or efficient paths that the
packets should follow

17
 Routers vs. bridges
Bridges (L2) are used for interconnecting LANs of the same type whereas routers (L3)
interconnect networks of different (or the same) types
• Bridge (+/-)
+ bridge operation is simple, requires less processing
+ transparent (no configuration needed when new nodes added to LAN)
– restricted topology (forwarding determined by a spanning tree)
– LANs use a flat addressing space (no hierarchical network structure)
• Router (+/-)
+ arbitrary topologies, enables use of efficient routing algorithms for
distributing traffic (helps traffic management)
+ hierarchical addressing enables scalability:
• scalability requires minimization of address info stored in routers
• routing based on network numbers  forwarding tables contain info
on
all networks, not all nodes
– requires IP address configuration
– packet processing more demanding
• Summary: bridges do well in small (~ 100 hosts) networks while routers
are used in large networks (1000s of hosts)
18
 IP forwarding
• Preliminaries
– Every datagram contains destination’s address
– Every router has a forwarding table
– Each host has a default router configured
– Routers maintain forwarding tables with multiple entries
(constructed via routing process)
– Forwarding table maps network number into next hop router
number or local interface number

• Strategy
– A router receiving a packet checks destination network address
of datagram and
– if directly connected to destination network, then forwards
directly to host
– need to map IP address to physical LAN address (ARP)
– if not directly connected to destination network, then forwards to
next hop router
19
 IP forwarding example

H1H3: forwarding on the same network

H5H8: forwarding via R1 and R2

Router R2

20
 Address translation
What to do when router/host notes that it is connected
directly to the network where an arriving packet is
destined?

• Need to map IP addresses into physical LAN addresses

– destination host
– next hop router

• Techniques
– encode physical LAN address in host part of IP
address  not scalable! (not implemented)
– table-based (maintain IP address, PHY address pairs)
ARP
21
 ARP details
• ARP (Address Resolution Protocol)
– utilizes LAN’s broadcast capabilities
– each node maintains table of IP to physical LAN address bindings
– broadcast request if address not in table
– target machine responds with its physical LAN address
• ARP request contains also source addresses (physical and IP)
– all “interested” parties can learn the source address
• Node (host/router) actions:
– table entries timeout in about 10 minutes
– if node already has an entry for source, refresh timer
– if node is the target, reply and update table with source info
– if node not target and does not have entry for the source, ignore
source info
22
 ARP packet format
– HardwareType: type of physical network (e.g., Ethernet)
– ProtocolType: type of higher layer protocol (e.g., IP)
– HLen & PLen: length of physical and upper layer addresses
– Operation: request or response
– Physical/IP addresses of Source and Target

23
 Need for automatic configuration!
• IP addresses need to be reconfigurable
– Ethernet addresses hardwired onto the network adapter
– IP address consists of network and host part
– hosts can move between networks host gets new address in each
network
• Need for automated host configuration
– hosts need other configuration info, e.g., the default router
– configuration manually impossible (too much work and errors)
– Dynamic Host Configuration Protocol (DHCP)
• DHCP server
– at least one DHCP server for each administrative domain
– centralized repository for configuration info
– two operation modes:
 administrator chooses host addresses and configures them to DHCP
 DHCP manages the addresses by allocating addresses dynamically
from a pool of available addresses (more sophisticated)

24
 DHCP operation
• Server discovery: host sends
DHCPDISCOVER msg to IP
broadcast address
(255.255.255.255)
• Msg broadcasted only on
same network
• If server on same network,
host receives its IP address
• If not, msg picked up by
DHCP relay agent
• Relay agent knows address of DHCP server, forwards the msg to DHCP
server and host receives its IP address
• Use of DHCP relay agent makes it possible to have fewer DHCP servers
(relay agent configuration simpler than DHCP server configuration)

25
 Internet Control Message Protocol
• ICMP used for reporting errors in Internet

• Messages
– Echo (ping)
– Redirect (from router to source host if router knows of
a better route to packet’s destination)
– Destination unreachable (protocol, port, or host)
– TTL exceeded (so datagrams don’t cycle forever)
– Checksum failed
– Reassembly failed
– Cannot fragment
26
 Virtual private networks (VPN)
• Problem:
– group of isolated networks
– geographically distant from each other
– need to connect different networks together into a
“private” network securely
– e.g., company with many branch offices
• Methods
– Build a private network from leased lines (not shared)
– Create a virtual private network (VPN) which is overlaid
on top of public networks
• ATM and Frame Relay: Use virtual circuits
• IP: Use IP tunneling (usually coupled with secure IPsec)

27
Private vs VPNs

28
 VPN and IP tunneling

• Problem with IP:

- not possible to connect to Internet via router without the whole Internet
also knowing about your network.
• Tunneling
– virtual point-to-point link between two nodes separated by arbitrary no of
networks
– created at R1 by providing it with address of R2
– R1 encapsulates original packet in a new packet addressed to R2
– packet forwarded normally inside IP network
– R2 receives packet and strips off packet header and notices payload
contains an encapsulated packet addressed to some host inside network 2
•Common use case: Tunnel IP over IP 29
 Global Internet

Scalability Issues
IP “hides” hosts in address hierarchy, but...
• Inefficient use of address space
– class C network with 2 hosts (2/255 = 0.78% efficient)
– class B network with 256 hosts (256/65535 = 0.39%
efficient)
• Too many networks
– today's Internet has tens of thousands of networks
– routing tables do not scale
– route propagation protocols do not scale
30
 Subnetting (1)

• The original intent of IP addresses was that the network

part would uniquely identify exactly one physical network
• Problem 1: Any network with more than 255 hosts,
needs class B addresses, or should get many class C
addresses
• Problem 2: Each new network implies additional entry in
forwarding table  large table
• Solution:
– Share one network number between several
networks.  another level is added to address/routing
hierarchy : subnet
31
 Subnetting (2)
• Makes most sense for large corporations or campuses
• Corporation networks share 1 network number
• Number of other networks within the corporation, using
subnet masks
– E.g., a class B address, is shared among 8 networks,
by using a 19-bit “subnet mask” (255.255.224.0 =
11111111 11111111 11100000 00000000)
– I.e., subnet addresses are defined by first 19 bits of
the IP address.  Host part now has a “subnet” part
in it.
• Class B network address continues to be advertised to
the rest of the Internet, subnetting only used “within
campus” (i.e. Subnets visible only within site).

32
 Subnetting (3)
• Another example: a class B address, is shared
among 256 networks, by using a 24-bit “subnet
mask” :
(255.255.255.0 = 11111111 11111111 111111111 00000000)

33
Subnet example

Forwarding table at R1:

34
 Forwarding algorithm with subnetting
D = destination IP address
for each entry <SubnetNum, SubnetMask, NextHop>
D1 = SubnetMask & D
if D1 = SubnetNum
if NextHop is an interface
deliver datagram directly to destination
else
deliver datagram to NextHop (a router)

Q: Apply the algorithm when H1 is sending to H2 and H3.

Notes
• Would use a default router if nothing matches
• Not necessary for all ones in subnet mask to be contiguous
• Can put multiple subnets on one physical network
• Subnets not visible from the rest of the Internet
35
 Subnetting Example
• Where does the router forward packets addressed to:
– 128.96.39.10 –> If0
– 128.96.40.12 –> R2
– 128.96.40.151 –> R4
– 192.4.153.17 –> R3
– 192.4.153.90 –> R4

Subnet Number Subnet Mask Next Hop

128.96.39.0 255.255.255.128 Interface 0
128.96.39.128 255.255.255.128 Interface 1
128.96.40.0 255.255.255.128 R2
192.4.153.0 255.255.255.192 R3
<default> R4
36
 Supernetting (CIDR) (1)
• Classful addressing:
– inefficient use of address space, address space exhaustion
– subnetting does not get around the fact that a network with more than
255 hosts requires a class B address
– class B net allocate addresses for 65K hosts, even if only 2K hosts in
that network
• CIDR: Classless InterDomain Routing
– network portion of address of arbitrary length
– address format: a.b.c.d/x, where x is # bits in network portion of address

network host
part part
11001000 00010111 00010000 00000000
200.23.16.0/23
37
 Supernetting (CIDR) (2)

• Assign block of contiguous network numbers to

near-by networks  helps to aggregate routes  less
entries in the forwarding tables
• Restrict block sizes to powers of 2
• E.g. Consider an Autonomous system with 16 class C
network numbers (192.4.16 to192.4.31) the top 20 bits
are same  20-bit network number (a single network
prefix) can be used in forwarding tables. Something
between class B and class C  classless!

38
 Route aggregation

To achieve scalability, you need

to reduce the amount of information
stored in routers
Hierarchical aggregation:
IP has a two-level hierarcy (network+host) 39
 A simple example
• Suppose hosts had arbitrary addresses
– Then every router would need a lot of information
– …to know how to direct packets toward the host

1.2.3.4 5.6.7.8 2.4.6.8 1.2.3.5 5.6.7.9 2.4.6.9

host host ... host host host ... host

LAN 1 LAN 2
router router router
WAN WAN

1.2.3.4
1.2.3.5

40
forwarding table
 Example continued..
• Number related hosts from a common subnet
– 1.2.3.0/24 on the left LAN
– 5.6.7.0/24 on the right LAN

1.2.3.4 1.2.3.7 1.2.3.156 5.6.7.8 5.6.7.9 5.6.7.212

host host ... host host host ... host

LAN 1 LAN 2
router router router
WAN WAN

1.2.3.0/24
5.6.7.0/24

forwarding table
41
 Example continued..
• Easy to add a new host
• No need to update the routers
– E.g., adding a new host 5.6.7.213 on the right
– Doesn’t require adding a new forwarding entry

1.2.3.4 1.2.3.7 1.2.3.156 5.6.7.8 5.6.7.9 5.6.7.212

host host ... host host host ... host

LAN 1 LAN 2
router router router host
WAN WAN
5.6.7.213
1.2.3.0/24
5.6.7.0/24

forwarding table
42
 Longest Prefix Match
• With CIDR, prefixes may be of any length and may
overlap
• For example, suppose we have the following in the
forwarding table at a router:
– 171.69 (a 16-bit prefix)
– 171.69.10 (a 24-bit prefix)
• A packet with destination address 171.69.10.5 matches
both prefixes
• The rule is to choose the longest prefix to forward to
• Consider also 171.69.20.5. The longest match would be
171.69
• Efficient algorithms are required for finding the longest
match in a forwarding table
43
 CIDR Example
• Suppose the router does the longest-prefix match. Where does the
router forward packets addressed (in hex) to:
– C4.5E.13.87 –> B
– C4.5E.22.09 –> A
– C3.41.80.02 –> E
– 5E.43.91.12 –> F
– C4.6D.31.2E –> C
– C4.6B.31.2E –> D
Net / Mask Length Next Hop
C4.50.0.0 / 12 A
C4.5E.10.0 / 20 B
C4.60.0.0 / 12 C
C4.68.0.0 / 14 D
80.0.0.0 / 1 E
40.0.0.0 / 2 F
44
00.0.0.0 / 2 G
 Today’s Internet
• Sites are connected arbitrarily rather than as a
tree-like structure as was the case in 1990s
• Some large corporations connect with each
other. This is called peering

45
 Today’s Internet and Cloud

• Cloud supports several warehouse-sized datacenters, each of which

provides a cost-effective way to power, cool, and operate a large number of
server machines
• End users connect to the nearest datacenter over the Internet
• Major cloud providers operate several dozen datacenters spread across the
globe, strategically located in close proximity to Internet Exchange Points
(IXP), each of which provides rich connectivity to the rest of the Internet
46
 Routing in the Internet (1)
• The Global Internet
consists of Autonomous
Systems (AS)
interconnected with
each other.
• AS:
– corresponds to an
administrative domain
– examples: University,
company, backbone
network
– Each AS is assigned
a 16-bit number

47
 Routing in the Internet (2)
• We can classify ASs into three types:
– Stub AS: small corporation: a single connection to
one other AS.  only carry local traffic
– Multihomed AS: large corporation: multiple
connections to other AS’s. (no transit, i.e. refuses to
carry others traffic)
– Transit AS: provider, hooking many AS’s together
– Q: Which ASes are Stub, Multihomed and Transit in
Slide 40?
• Two-level routing:
– Intra-AS: administrator responsible for choice of
routing algorithm within network
– Inter-AS: unique standard for inter-AS routing: BGP

48
 Intra-AS, Inter-AS routing
• Most common Intra-AS routing protocols:
- RIP: Routing Information Protocol (Distance-Vector Routing)
- OSPF: Open Shortest Path First (Link-State Routing)
- Find the lowest-cost paths
• Most common Inter-AS routing protocol is Border Gateway Protocol
(BGP)
- Leave “optimality” aside. Just find a loop-free path
- Thus only bothers with reachability
R4
R5 BGP
R3 AS3
(OSPF intra-AS
AS1 AS2 routing)
(RIP intra -AS (OSPF
routing) BGP intra-AS
routing)
R1 R2
49
Routing: optimality vs scalability
• Distance-vector and link-state are both distributed
routing algorithms
• In distance-vector, each node talks only to its directly
connected neighbors, but it tells them everything it has
learned (i.e., distance to all nodes)
• In link-state, each node talks to all other nodes, but it
tells them only what it knows for sure (i.e., only the state
of its directly connected links)
• Inter-domain routing: In large networks, scalability is a
more pressing design goal than selecting the optimal
route

50
 Software-Defined Networks
• Disaggregation: Decouple the network control plane (i.e.,
where routing algorithms like RIP, OSPF, and BGP run)
from the network data plane (i.e., where packet
forwarding decisions get made), with the former moved
into software running on commodity servers and the
latter implemented by white-box switches
• The key enabling idea: take this decoupling a step
further and to define a standard interface between the
control plane and the data plane
– Doing so allows any implementation of the control
plane to talk to any implementation of the data plane;
this breaks the dependency on any one vendor’s
bundled solution (OpenFlow, the P4 language) 51
 “Centralized” control plane
• Another important aspect of disaggregation is that a
logically centralized control plane can be used to control
a distributed network data plane
• We say logically centralized because while the state
collected by the control plane is maintained in a global
data structure, such as a Network Map, the
implementation of this data structure could still be
distributed over multiple servers. For example, it could
run in a cloud.
• This is important for both scalability and availability,
where the key is that the two planes are configured and
scaled independently of each other
52
Network operating system

53
 IPv6 features

Major Features
• 128-bit addresses
• Multicast
• Real-time service
• Authentication and security
• Autoconfiguration
• End-to-end fragmentation
• Protocol extensions

54
 IPv6 addresses
IPv6 addresses
• Classless addressing/routing (similar to CIDR)
• Notation: x:x:x:x:x:x:x:x (x = 16-bit hex number)
– contiguous 0s are compressed:
47CD::A456:0124
– IPv6 compatible IPv4 address: ::128.42.1.87
• Address assignment
– provider-based
– geographic

55
 IPv6 Header
• 40-byte “base” header
• Extension headers
(fixed order, mostly
fixed length)
– fragmentation
– source routing
– authentication and
security
– other options

56
 Transition to IPv6
• IPv6 needs to be deployed incrementally in such a way
that hosts and routers that only understand IPv4 can
continue to function for as long as possible
• Two major mechanisms have been defined to help this
transition: dual-stack operation and tunneling
• Dual stack: IPv6 nodes run both IPv6 and IPv4 and use
the Version field to decide which stack should process
an arriving packet
• Tunneling: Sending an IP packet as the payload of
another IP packet
– For IPv6 transition, tunneling is used to send an IPv6
packet over a piece of the network that only
understands IPv4 57
 Switch/Router implementation
• Implementations of switches and routers have
so much in common that a network administrator
typically buys a single forwarding box and then
configures it to be an L2 switch, an L3 router, or
some combination of the two
• We have software switches and hardware
switches
– Both approaches obviously include a
combination of hardware and software

58
 Software switches

• Two potential bottlenecks limit the aggregate packet forwarding capacity of

the software switch: Memory read/write speeds and cost of processing each
packet (especially when packets are small) 59
 Hardware switches
• High-performance switches and routers have
been specialized devices, built with Application-
Specific Integrated Circuits (ASICs)
– High required throughput rates achievable
– Problem with ASICs is that hardware takes a
long time to design and fabricate, meaning
the delay for adding new features to a switch
is usually measured in years
• Closed “black-box” devices have historically
dominated the industry
60
 Hardware switches
• Recent advances in domain-specific processors (and other
commodity components) have made it possible to build switches
with performance of ASICs and while benefiting the agility of
software
Network Processor Unit (NPU):
a domain-specific processor
with an architecture and
instruction set that has been
optimized for processing packet
headers
Ternary Content Addressable
Memory (TCAM): key you want
to look up in a table can
effectively be used as the
address into the memory
that implements the table
61